Date of Award

2006

Degree Type

Thesis

Degree Name

Master of Engineering Science

Program

Electrical and Computer Engineering

Supervisor

Dr. Jagath Samarabandu

Abstract

The capability of identifying physical structures of an unknown environment is very important for vision based robot navigation and scene understanding. Among physical structures in indoor environments, corridor lines and doors are important visual landmarks for robot navigation since they show the topological structure in an indoor environment and establish connections among the different places or regions in the indoor environment. Furthermore, they provide clues for understanding the image. In this thesis, I present two algorithms to detect the vanishing point, corridor lines, and doors respectively using a single digital video camera. In both algorithms, we utilize a hypothesis generation and verification method to detect corridor and door structures using low level linear features. The proposed method consists of low, intermediate, and high level processing stages which correspond to the extraction of low level features, the formation of hypotheses, and verification of the hypotheses via seeking evidence actively. In particular, we extend this single-pass framework by employing a feedback strategy for more robust hypothesis generation and verification. We demonstrate the robustness of the proposed methods on a large number of real video images in a variety of corridor environments, with image acquisitions under different illumination and reflection conditions, with different moving speeds, and with different viewpoints of the camera. Experimental results performed on the corridor line detection algorithm validate that the method can detect corridor line locations in the presence of many spurious line features about one second. Experimental results carried on the door detection algorithm show that the system can detect visually important doors in an image with a very high accuracy rate when a robot navigates along a corridor environment.

Share

COinS
 
 

To view the content in your browser, please download Adobe Reader or, alternately,
you may Download the file to your hard drive.

NOTE: The latest versions of Adobe Reader do not support viewing PDF files within Firefox on Mac OS and if you are using a modern (Intel) Mac, there is no official plugin for viewing PDF files within the browser window.